Operational performance assessment of additive manufacturing

Background

Additive manufacturing is a process by which a three-dimensional structure is built, usually in a series of layers, based on a digital model of the structure. The process is sometimes referred to as three-dimensional (3D) printing or 3D rapid prototyping, and the term “print” is often used even though some examples of the technology rely on sintering or melting/fusing by way of an energy source to form the structure, rather than “printing” in the traditional sense where material is deposited at select locations. Examples of additive manufacturing techniques include Fused Deposition Modeling, Electron Beam Melting, Laminated Object Manufacturing, Selective Laser Sintering (including Direct Metal Laser Sintering, also referred to as Direct Metal Laser Melting, also referred to as Selective Laser Melting), and Stereolithography, among others. Although 3D printing technology is continually developing, the process to build a structure layer-by-layer is relatively slow, with some builds taking several days to complete.

One of the disadvantages of current additive manufacturing processing relates to quality assurance. There is typically some amount of analysis to determine whether the produced part meets the manufacturing thresholds and design criteria. In some examples, the parts can be evaluated using non-destructive engineering, such as scanning, to ensure that the part meets the design thresholds. However in other cases, the part may have to be dissected in order to test whether a certain lot of products or a sampling has satisfied the design limits. This can lead to considerable inefficiency when, for example, it is later determined that a production lot is defective due to a machining or design problem.

There have been some attempts to alleviate the aforementioned problem. In one example, for selective laser sintering, images are obtained to provide a crude estimation of the production process for the large features. A scoring system is utilized to determine if a part fails. However, such a system is unable to determine the root cause analysis of the failure. In the traditional 3D printing area, there are currently limited techniques.

For at least the reasons recited, there is a growing need for real-time inspection systems and processes that can evaluate the additive manufacturing products and assess the quality of the products and health of the systems.

Brief description

Assurance that a build process is progressing to plan can be important, given the resources, both in time and material, that are expended. In accordance with aspects described herein, a method is provided for assessment of operational performance of a 3D manufacturing apparatus. The method includes, for instance: obtaining, in real-time during a 3D polymer printing build process in which at least one structure is built by the 3D manufacturing apparatus, images of an area of a build platform on which the at least one structure is built; evaluating, by a processor, the obtained images; and determining, based on the evaluating, whether an operational flaw with the 3D manufacturing apparatus has occurred.

Additionally, a system is provided for assessment of operational performance of an additive manufacturing apparatus. The system includes, for instance: a memory; and a processor in communication with the memory, wherein the system is configured to perform: obtaining, in real-time during an additive manufacturing build process in which at least one structure is built by the additive manufacturing apparatus, images of an area of a build platform on which the at least one structure is built; evaluating, by a processor, the obtained images; and determining, based on the evaluating, whether an operational flaw with the additive manufacturing apparatus has occurred.

Further, a computer program product is provided for assessment of operational performance of a 3D manufacturing apparatus. The computer program product includes, for instance: a non-transitory computer readable storage medium readable by a processor and storing instructions for execution by the process to perform a method comprising: obtaining, in real-time during a 3D polymer printing build process in which at least one structure is built by the 3D manufacturing apparatus, images of an area of a build platform on which the at least one structure is built; evaluating the obtained images; and determining, based on the evaluating, whether an operational flaw with the 3D manufacturing apparatus has occurred.

Additional features and advantages are realized through the concepts of aspects of the present invention. Other embodiments and aspects of the invention are described in detail herein and are considered a part of the claimed invention.

Drawings

One or more aspects of the present invention are particularly pointed out and distinctly claimed as examples in the claims at the conclusion of the specification. The foregoing and other objects, features, and advantages of the invention are apparent from the following detailed description taken in conjunction with the accompanying drawings in which:

FIG. 1 depicts one example of an additive manufacturing apparatus, in accordance with aspects described herein;

FIG. 2 depicts an image of an area of a build platform having multiple structures printed thereon and exhibiting an operational flaw with the additive manufacturing apparatus, in accordance with aspects described herein;

FIGS. 3A & 3B depict images of overhead and angled-overhead views of a witness fence printed by an additive manufacturing apparatus to facilitate operational performance assessment of the additive manufacturing apparatus, in accordance with aspects described herein;

FIG. 4 depicts an example process for building one or more structures using the additive manufacturing apparatus of FIG. 1 , in accordance with aspects described herein;

FIG. 5 depicts another example of an additive manufacturing apparatus, in accordance with aspects described herein;

FIGS. 6A-6D depict alternative arrangements for the camera component of the additive manufacturing apparatus of FIG. 5 , in accordance with aspects described herein;

FIGS. 7A-7B depict images of angled-overhead views of an area of a build platform having a powder bed disposed thereon and exhibiting operational flaws with the additive manufacturing apparatus, in accordance with aspects described herein;

FIG. 8 depicts an example process for building one or more structures using the additive manufacturing apparatus of FIG. 5 , in accordance with aspects described herein;

FIG. 9 depicts an example of a process for assessment of operational performance of additive manufacturing during an additive manufacturing build process, in accordance with aspects described herein;

FIG. 10 depicts one example of a data processing system to incorporate and use one or more aspects described herein; and

FIG. 11 depicts one example of a computer program product to incorporate one or more aspects described herein.

Detailed description

The phrase “additive manufacturing apparatus” is used interchangeably herein with the phrase “printing apparatus” and term “printer”, and the term “print” is used interchangeably herein with the word “build”, referring to the action for building a structure by an additive manufacturing apparatus, regardless of the particular additive manufacturing technology being used to form the structure. As used herein, print and printing refer to the various forms of additive manufacturing and include three-dimensional (3D) printing or 3D rapid prototyping, as well as sintering or melting/fusing technologies. Examples of additive manufacturing or printing techniques include Fused Deposition Modeling, Electron Beam Melting, Laminated Object Manufacturing, Selective Laser Sintering (including Direct Metal Laser Sintering also referred to as Direct Metal Laser Melting, also referred to as Selective Laser Melting), and Stereolithography among others.

Assurance that a build process is progressing as planned is important for cost and quality reasons. At the end of a build cycle to build one or more three-dimensional structures, an operator of the additive manufacturing apparatus may find that the parts are defective or unusable because of a failure with the additive manufacturing apparatus during the build cycle. This can be especially problematic when building expensive parts, such as molds for casting structures having complex geometries.

A machine vision-based process monitoring system is disclosed herein that may be used to monitor the building of layers of one or more objects being built by an additive manufacturing apparatus, and, in one embodiment, to detect operational flaws as they occur, i.e. during the build process rather than afterward, as an example. In a further embodiment, evaluation/analysis of images acquired during the build process is performed as part of post-processing (and not as part of the real-time acquisition of images). Real-time acquisition as used herein refers to the image capture of individual layer(s) of the structure as the structure is being built (“printed”). Real-time analysis refers to evaluation of the acquired images of the various layers.

Operational flaws may include, as examples, errors with the structure(s), build process, or additive manufacturing apparatus, or indicators that one or more errors are likely to occur with the structure(s), build process, or additive manufacturing apparatus. In some embodiments, action(s) may be taken responsive to observing that an operational flaw has occurred. For instance, remedial actions may be taken so that the flaw can be corrected, the build process stopped, the problem fixed, a new build started, etc.

Provided is an ability to observe a build process that may take hours or days to complete in order to detect and react to potential operational flaws with the additive manufacturing apparatus and/or errors with one or more printed layers. Also provided is the ability to communicate indications of the operational flaws to operators early in the build process as, or before, they occur, so that a failed build can be stopped prior to its completion. A new build may then be started earlier than it otherwise would have been (i.e. had the failure been discovered only after the failed build process completes). From a manufacturing resources perspective, wasted materials usage and wasted build time are reduced. In addition, as described below, rather than stopping an entire build process, printing of individual parts that are showing flaws or otherwise undesired features can be turned off so as the flaws/features do not cause the build to fail, which could cause errors with all of the structures in the build. By terminating building of individual parts that are becoming problematic, manufacturing yields and machine uptime can be maximized.

Some problems that may be observed during the monitoring of a build process as described herein include, but are not limited to, dimensional errors, distortion, or cracking in the printed structures, failed or clogged print jets (nozzles), malfunctioning of a roller/planarizer or other component of the printing apparatus, poor layer surface finish, delamination of the structures, misplacement, excess, or absence of build material, or any other additive manufacturing errors. In general, the monitoring can monitor for anything that can cause the built part to fail or that can indicate that that additive manufacturing apparatus has failed, is about to fail, or needs maintenance, as examples.

An example additive manufacturing apparatus and associated process in accordance with aspects described herein are presented with reference to FIGS. 1-4 , in the context of printed disposable dies for casting of parts. The disposable core dies in this example are built out of printed polymer material, though other materials are possible.

In one example, the printing apparatus prints the structures in layers. For the first layer, a print head moves across a build platform and polymer is released onto the build platform at only the correct ‘pixel’ locations for that layer. A light source with an appropriate wavelength is then passed over the portion that was printed, curing it in place. After this layer is complete, the build platform lowers a distance that is equal to the layer thickness of the build (this is usually predetermined by the operator of the system). Then, the next layer is printed on top of the previous layer, and this cycle continues. It should be noted that such a 3D polymer printing process involving a print head laying down layers on a build platform is distinguished from other forms of additive manufacturing, such as powder-based laser sintering, described in further detail below.

A disposable die may be a very complex structure, having layers in which sections of polymer would be unsupported if printed alone. There may be areas of polymer within a layer that are difficult to print in that there is nothing underlying that area of the layer (i.e. no polymer was printed at that area of the preceding layer). The result is that there is no underlying structure onto which the polymer at that location may be printed. It is not possible to print polymer in mid-air, and so the printer may be configured to print a different material as a support material for the polymer to be printed in subsequent layer(s). This different material may be wax, for example, through wax is not the only kind of support material that could be used. In this case, the printer may print both polymer and wax concurrently, printing the wax only where no polymer is being printed but where a supporting structure is needed for material in an upper layer. When the print process to print the polymer structure is complete, the final as-printed structure may be the polymer structure at least partially encased in wax support material. The wax may then be melted away, leaving the polymer structure behind.

One potential challenge in the above process is distortion of the printed structure. If there are errors in printing—for instance holes in the polymer walls, cracks, or shape distortions, as examples—then the printed structure may not function as intended in its downstream application.

By way of some examples, holes in the polymer structure may be the result of clogged print head jets. The holes may be hard to see with the naked eye but can cause cast material to leak during the casting process, resulting in significant losses in production yields and production time (clean up, refixturing, etc.). Walls on the polymer structure that are too thin can lead to bulging and distortion during the casting process. Furthermore, mixing between the polymer and wax materials (in this example) at the boundaries between them when printed can lead to porosity in the polymer structure and increased surface roughness. This can impact the strength of the polymer die and change characteristics of the casting process, leading to distorted cast structures. The above problems and others may lead to manufacturing failures that may be extremely expensive, for instance when they cause defects in expensive parts being cast from the printed dies. In relation to 3D polymer multijet printing processes, there are no known imaging systems to detect defects on the layer level.

According to aspects described herein, an imaging system is leveraged for monitoring of build quality and machine health during an additive manufacturing process to build a structure, so that the quality of the structure being built and the health of the additive manufacturing apparatus can be assessed. Aspects of the monitoring and analyzing can be performed in real-time, e.g. during the build process. The monitoring includes, in some embodiments, capturing images of the build during the build process (real-time acquisition of images of the build process). This may include, for instance, images of area(s) of the build platform, including images of the individual layers of the structure(s) as the layers are being built, images of one or more additive manufacturing apparatus components, etc., as examples. An assessment of part quality and machine health may then be performed by evaluating the images. For instance, the captured images may be evaluated to ascertain characteristics (dimensions, textures, composition, etc.) of the structure(s) being printed and compare these to a ‘golden standard’, such as a computer-aided design (CAD) specification for the structure. The CAD specification may be a specification that the additive manufacturing apparatus uses in building the structure. The comparison can assess whether the structure is being built consistent with the CAD specification in order to identify possible distortions, deviations, or other flaws.

Since, build quality is dependent on machine performance, the evaluation of the images can additionally identify features in the images that suggest problems with the additive manufacturing apparatus, such as, in this example in which nozzles deposit print material, a clogged print head nozzle, improperly functioning roller or planarizer, improper surface finishing, or any other observable items that indicate a flaw. Thus, the images can be evaluated to not only detect errors in the structure(s) being built as they are printed, and assign a part ‘health’ score to the structure(s), but also monitor additive manufacturing apparatus health, indicating when the machine might require maintenance and identifying what is needed for that maintenance. In some examples, the evaluation is performed in real-time during the build process, though in other examples, the evaluation is performed at a later time.

The imaging system or other system for evaluating the images can include customized camera control software and customized defect detection software developed using, for instance, the LabVIEW® graphical development environment (LabVIEW® is a registered trademark of National Instruments Corporation, Austin, Tex.).

When the evaluation of the acquired images reveals a problem, one or more actions may be taken in response, and the types of actions may vary. For instance, an operator of the additive manufacturing apparatus may be notified of the problem. In some embodiments, an auditory or visual alarm or alert, or an electronic communication (i.e. text or email), is provided to the operator indicating that the flaw has occurred. Additionally or alternatively, adjustments may be made to the additive manufacturing process. The process may be halted for instance. In this regard, some errors may be not recoverable, necessitating shut down of the machine in order to allow for operator intervention. However, in some instances, such as if the error is exhibited only when building a particular part or row of parts, the process is modified but not halted altogether; instead, the process is optionally continued to a next phase, skipping the building of object(s) where the operational flaw(s) is/are exhibited. For instance, a ‘bad row’ of parts or problematic area of the build platform may be noted and the rest of the build may be completed. Noting the bad row may include notifying the operator of the bad row of parts. In further embodiments, the build process may be continued despite observing occurrence of an operational flaw, and, if the error occurs over a substantial area of the build platform or with a threshold number of parts, then the rest of the build may be halted.

Detection algorithms can be used in the evaluation of the acquired images in order to detect the built structure(s), compare them to the CAD model, and identify distortions or deviations in the build structure(s). Early detection of operational flaws may reduce manufacturing time spent on failed part builds, reduce scrap, reduce raw materials usage, and increase up time on additive manufacturing equipment, as examples.

Some of the failure modes identified above may be observable from images taken of the useful parts during building thereof. However, due to geometries of the useful parts or other reasons, it may be difficult to observe some or all of the above failure modes by observation of the useful parts alone. Accordingly, one or more additional structure(s) may be printed to assist in monitoring the health of the additive manufacturing apparatus as the useful parts are built. An example such structure is a calibration target. Another example is a witness fence (also termed a build fence), which is, in one example, an extraneous printed structure not intended for further use beyond its purpose of serving as another observable structure during the print process.

The witness fence may be specially designed into the CAD specification so that each material being printed may have a corresponding portion of the witness fence. In other examples, each material has a corresponding individual witness fence for the material. In this manner, the witness fence becomes another part to be printed at the same time as the other parts. In addition, or as an alternative, to imaging the useful parts being printed, light may be focused on the witness fence, and the light reflecting off of the witness fence may be imaged by one or more cameras. Variations in the reflected light may be identified to indicate operational flaws with the printer, such as a problem with the print head. Also, rough textures on the witness fence may be identified to indicate a problem with the roller(s) or planer(s) of the printer.

One or more cameras may be dedicated to imaging the witness fence, and one or more other cameras may be dedicated to imaging other portions of the build area (i.e. the area where the useful parts are being printed). Alternatively, the witness fence and useful parts may be imaged by the same one or more cameras. Additionally, while aspects are described herein in the context of lights and cameras for process monitoring, other forms of optical and/or acoustic sensing, such as linear photodiode array or ultrasound imaging technologies, are also possible. In one example, high resolution imaging is utilized such that particle size images are obtained. As used herein, the term particle size images refers to high resolution images where the optical resolution is at least on the order of the particle size of the powder used in the process. In one example particle size images refers to optical resolution greater than the particle size enabling sub particle imaging.

FIG. 1 depicts one example of an additive manufacturing apparatus, in accordance with aspects described herein. As is seen in FIG. 1 , printing apparatus 100 is a 3D printing device that includes a print head 102 mounted to arm 104 . Arm 104 drives the positioning of print head 102 over build platform 110 , onto which material is deposited. Print head 102 may include a plurality of print nozzles through which material, e.g. polymer and wax material, is deposited onto build platform 110 . Also mounted to arm 104 are line scan camera 106 and light source 108 (in this example a light emitting diode (LED) light source). While a line scan camera is illustrated in this example, other imaging devices can be utilized as detailed herein.

Light source 108 is a light that can be focused into a tight line that can be position so that it overlaps the imaging location of the line scan camera, giving a uniform illumination field for the line scan camera. In one example, printing apparatus 100 is a modified version of a commercially available 3D printer, modified to include one or more cameras and one or more light sources. Example such commercially available 3D printers include those of the ProJet® line of printers offered by 3D Systems Inc., Rock Hill, S.C. (ProJet® is a registered trademark of 3D Systems, Inc.).

Printing apparatus 100 may also include a control system including one or more controller(s) 112 , including hardware and/or software for controlling functioning of some or all components of printing apparatus 100 . Controller(s) 112 may control, for instance, positioning of print head 102 and/or deposition of materials therefrom, functioning of light 108 , and/or functioning of camera 106 . In some embodiments, controller(s) 112 include one or more control data processing systems for controlling the print process and behavior of the other hardware of the printing apparatus.

With respect to the mounted camera(s) 106 and light(s) 108 , the lighting may be developed and positioned to highlight the particular features of the printed structures (i.e. the useful parts or the witness fence) that are of interest. The camera(s) may be mounted in the vicinity of the build platform, such as an internal camera within the build chamber, and mounted for instance on the arm to which the print head is attached, as in FIG. 1 . Additionally or alternatively, camera(s) and/or lighting may be mounted about or adjacent to a build chamber. In some examples, the camera is an external camera that views the build process from outside of the build chamber through, an aperture, door, or window, as examples. Example alternative configurations are depicted in FIGS. 6A-6D .

Various types of cameras may be used. In general, line scan cameras can produce very high resolution images, enabling detection of features that would otherwise go undetected with lower resolution equipment. Many line scan cameras are capable of producing images having resolution of 12 K at 50 um per pixel, though even small pixel width of 5-15 um is possible in some cases. Line scan cameras, however, need to be moved over the area to be imaged. Other types of cameras, such as those of the Digital Single-lens Reflex (DSLR) type, do not need to be moved, can more easily sit outside or stationarily within the build chamber, and can capture images at an angle with specialized lenses. Image resolution, however, is generally not as robust as that of high-end line scan cameras. Accordingly, the particular imaging equipment used may vary depending on the circumstances and desired flaws to monitor.

The camera(s) may capture images in real-time during the build process. The images may then be evaluated, in real time, in one example, using one or more algorithms executed as software on a data processing system. The data processing system may be included as part of the camera, in one example. In other examples, the data processing system is in wired or wireless communication with a camera responsible for acquiring the images, where the camera communicates the images through one or more wired or wireless communication paths to the data processing system. The separate data processing system may be a controller ( 112 ) data processing system described above, or may be a different data processing system dedicated to evaluation of the acquired images.

In any case, the data processing system that obtains the images may evaluate the images, either alone or by one or more of various techniques for comparison with one or more 3D CAD models, to determine whether the structure(s) are being printed correctly. In a typical build setup, a designer of the structures to be printed may utilize software to build designs for all of the parts to be printed onto the build platform. Software for controlling the additive manufacturing apparatus may then (offline) ‘slice’ the 3D models of the structure(s) to be printed into layers, with each layer to be printed as a ‘pass’ of the print head. In embodiments that include printing of a witness fence, a design for the witness fence may be added by the designer to the CAD file's distribution of parts. In other embodiments, the design for the witness fence may be automatically added to the CAD file absent designer input, and/or the printer may be configured to automatically print a witness fence when other structures are printed. The witness fence model may be sliced into layers along with the other parts, and printed together with those parts.

As described above, in one example, the additive manufacturing apparatus prints two (or more) materials, and in the example of FIGS. 1-4 , prints wax and polymer. The particular lighting and imaging techniques used may be tailored based on these materials and the particular structure(s) being printed so that the imaging can most clearly show the materials being used. This can facilitate the detection/identification of the materials and the comparison to the CAD model. The lighting mode used for the useful parts may be different from the lighting mode used for the witness fence. In this manner, different lighting characteristics (orientation, intensity, etc.) may be utilized in order to detect the appropriate features on the different structures. Similarly, lighting for imaging the useful parts as they are built may be tailored around the particular defect(s) desired to be detected. In some examples, standard image processing tools are utilized to perform the evaluation of the images. Regarding the imaging techniques, confocal optics may be tailored to limit the depth of field for the sensing system, eliminating light above the image plane, which can help reduce the effect of sparks and light from plasma that may hover over a melt pool. This may substantially reduce noise and help to improve signal-to-noise ratio.

In the example of FIG. 1 , the printer's print head includes one or more rows of jets (nozzles) that print curable material (polymer in this example). An ultraviolet light follows behind these nozzles to cure the material as it is printed. The printer's print head also includes one or more rows of heated jets that print the support material (wax in this example). As the print head traverses the build platform, the control system determines when to activate the deposition of material from the jets. The two materials are deposited nearly simultaneously, where the delay between these two materials being printed is based on the speed of the print head and the distance between the two rows of jets.

The camera(s) mounted onto or about the additive manufacturing apparatus can include one or more witness fence cameras and one or more useful part cameras, which may or may not be the same set of cameras. The cameras can intermittently or periodically acquire still frame images of the build process and/or video of the build process. In some examples, the resolution of the cameras is set to about 25 micrometers per pixel, corresponding approximately to a 2″ field of view, though cameras with different resolutions and/or different fields of view may alternatively be used.

A line scan camera is used in the example of FIG. 1 . As the print head moves across the build platform depositing material(s), the line scan camera images a single line of pixels, then a next line of pixels, and so on, continuously and very rapidly, and then combines the lines of pixels together to form an image. Additionally or alternatively, area cameras may be used instead of line scan cameras to acquire the images, if desired. In general, different materials will appear differently in the obtained images, thereby enabling a data processing system to evaluate the images.

As noted above, not only may the build of the useful parts be imaged, but the build of the witness fence, in these examples, may also be imaged. The witness fence includes, in one example, adjacent and alternating portions of each material being printed. The witness fence may be imaged and the images evaluated to detect, for instance, clogged print nozzles and/or material mixing. By imaging this witness fences, clogged/failed printer nozzles can be detected and the operators alerted of a problem.

The printer may be configured to build the witness fence(s) as the printer prints rows of material. The witness fence(s) are built vertically, like the other structures being printed. In some embodiments, a witness fence for each material printed spans the entire width of the build platform. The witness fences may be imaged from one or more angles as they are printed, and these images evaluated. Evaluation of the images of the witness fence can include determining whether there are gaps in porosity of the layers of the witness fence. Additionally, air bubbles may be identified. The images can identify areas where the nozzles for a material are printing correctly or whether they are clogged. In examples in which polymer and wax are printed, a witness fence may include of a section of wax, then a section of polymer, then alternating thin sections of wax-polymer-wax-polymer, etc., so that the boundary regions between the printed polymer and wax may be examined to identify mixing of the two materials. Thus, the imaging of the witness fence and evaluation of those images provides another way to assess print quality and additive manufacturing apparatus health.

An example is now described with reference to FIG. 2 that illustrates an overhead image of an area of a build platform having multiple structures printed thereon and exhibiting an operational flaw with the additive manufacturing apparatus, in accordance with aspects described herein. The overhead image of FIG. 2 depicts a portion of witness fence 202 and portions of useful parts 204 and 206 being printed. The printed parts 204 , 206 include, in this layer, wax material (lighter, thicker rings) separating polymer portions (darker, thinner rings).

Witness fence 202 extends horizontally across the top portion of the image in FIG. 2 , and includes alternating walls of polymer and wax, with the brighter portions being printed wax material, and the darker portions being printed polymer material. The brightest (white) spots indicate where the jets have failed, i.e. spots where a polymer jet clogged and no (or too little) material was printed. The white spots are also visible on printed part 204 . Failed jets are just an example of one operational flaw that may be detected by observing the build of the witness fence. Other detectable printing failure mechanisms include misaligned planarizer, wrong planarizer speed, mixing or materials, and wrong volume of material being printed, as examples.

FIGS. 3A & 3B depict images of another example of a witness fence printed by an additive manufacturing apparatus to facilitate operational performance assessment of the additive manufacturing apparatus, in accordance with aspects described herein. FIG. 3A depicts an image of an overhead view of a witness fence. The image of FIG. 3A was taken looking straight (orthogonally) down on the fence, with light incident at 45 degrees from the vertical in the machine direction. FIG. 3B depicts an angled-overhead view of the same witness fence. The image of FIG. 3B was taken using a specular light and the camera held at approximately a 45° angle from vertical.

Referring to FIG. 3A , the witness fence extends from top to bottom (in this image) and includes alternating strips of wax 302 a , 302 b , 302 c , 302 d , and polymer 304 a , 304 b , 304 c , 304 d . In this example, the witness fence includes (traversing from left to right) a relatively thick strip of wax and relatively thick strip of polymer, then alternating thinner strips of wax-polymer-wax-polymer-wax-polymer.

The white spotting seen in the area marked 306 indicates clogged polymer jets. Also observable in area 306 is a discontinuity in the bright white dashed line separating the polymer strip 304 a and thinner wax strip 302 b , indicating that, for instance, mixing or insufficient deposition is occurring at the interface between those two printed materials.

Referring to FIG. 3B , when the light reflecting off the witness fence is imaged, shadows will appear at areas where jets failed to print material. Accordingly, area 306 is clearly noticeable as an area of clogged polymer jets, as above.

Area 308 in FIGS. 3A & 3B depicts an engineered gap in the witness fence for testing purposes. This engineered gap may enable development of detection algorithms to detect similar gaps in the witness fence as well as to optimize lighting to optimize (maximize in one example) the contrast of these features. As seen in FIGS. 3A and 3B , the contrast and visibility of these engineered gaps is higher with the specular lighting used in FIG. 3B than with the dark-field lighting used in FIG. 3A .

Both the overhead and the angled-overhead imaging modalities have their own advantages in terms of presenting defects and other indicators of operational flaws in the printing apparatus.

As described herein, layers of a build process may be imaged and the properties and characteristics of the printed materials may be compared to a CAD specification in order to assess the quality of the build and determine whether operational flaw(s) have occurred. The imaging of one or more layers in real time during the additive manufacturing process, and the evaluation of the images, which may be in real-time during the build process or may be at a later time, provides online inspection and process monitoring that facilitates assessment of the operational health of the additive manufacturing apparatus.

FIG. 4 depicts an example process for building one or more structures using the additive manufacturing apparatus of FIG. 1 , in accordance with aspects described herein. The process begins by positioning the stage (build platform) for building the next layer ( 402 ), which initially is the first layer of the build. Material for that layer is deposited on the stage, per the specifications of the model of the structure(s) being printed ( 404 ). During and/or subsequent to the deposition of material for that layer, the structure(s) being printed are imaged ( 406 ) to generate images of the build. It is determined, based on evaluation of those images, whether an operational flaw has occurred ( 408 ). As noted above, this determination accounts for both manifested flaws, as well as indication(s) that a flaw may, or is likely to, manifest itself later in the build process.

If no operational flaw is determined to have occurred, the process continues with a determination of whether the job is complete ( 410 ). If the job is not complete, meaning additional layer(s) are to be printed, then the process returns to ( 402 ) to position the stage for the next layer to be printed. If it is determined that the job is complete, then results of the build may be reported ( 412 ). In one example, a report or other indication of job completion is provided to a machine operator or other entity.

If at ( 408 ) it was instead determined that an operational flaw has occurred, the process determines whether the exhibited flaw renders the layer (as a whole or where the flaw is exhibited) too flawed to be used ( 414 ). In some examples, an operational flaw may be one that places the layer (or portion thereof) out of specification but which is nonetheless acceptable for use, at least at this point in the process. For instance, a minor deformity in a structure may be observed, but it may be deemed an acceptable variance. If the layer is not too flawed, the process continues to ( 410 ) to determine whether the job is complete.

If instead it is determined that the layer is too flawed, remedial action(s) may be taken. In this example, a determination is made as to whether to reprint the problem area ( 416 ). Reprinting the problem area may be helpful in situations where too little material was deposited at a particular location, which may be the result of a temporarily clogged print nozzle. If it is determined that the problem area is to be reprinted, the reprinting is performed ( 418 ), which in some examples necessitates a dynamic adjustment to the print process to effectuate the reprinting of that portion. If reprinting occurs, the process returns to ( 406 ) to image the printed structure again and determine whether any operational flaws continue to be exhibited ( 408 ).

The description continues in the full USPTO document.